AI-ASSISTED DRUG DISCOVERY PATENTS: FROM ALGORITHM TO ASSET
A familiar scene is playing out in life sciences companies. An AI platform screens millions of compounds, proposes a handful of promising structures, and the team sees a potential breakthrough. Someone sends a quick message to in-house counsel:
“The model designed this series. Can we still get a patent – and who is the inventor?”
The question is no longer hypothetical. A 2024 patent landscape report analyzed 1,087 AI-driven drug discovery patents, with the United States leading and China close behind, underscoring how quickly this area is filling with prior art and competition in AI patent inventorship. Patent analytics show AI-related pharma filings are growing at around 23% per year in the early 2020s.
At the same time, studies of AI-native drug developers have raised concerns about thinner biological data and earlier patent filings, which can affect both patent quality and regulatory strategy.
Against that backdrop, getting inventorship wrong is not a technicality. It can undermine enforceability, complicate deals, and invite challenges from competitors.
The Legal Baseline: AI Is a Tool, Not an Inventor
Across major patent offices your teams care about, rules on AI in drug discovery patents have converged on one point: AI systems cannot be named as inventors, only natural persons can.
In the United States, the Federal Circuit’s decision in Thaler v. Vidal confirmed that under the Patent Act, an “inventor” must be an “individual,” and “individual” means a human being. Therefore, an AI system such as DABUS (the central focus of the Thayer decision) cannot be listed as an inventor on a patent application.
Following that decision, the USPTO issued Inventorship Guidance for AI-Assisted Inventions in 2024 and updated it again in November 2025, reiterating that AI-assisted inventions are evaluated under the same legal standard as any other invention.
The same legal standard for inventorship applies whether or not AI is involved.
AI systems are treated as tools used by human inventors.
Only natural persons who make a significant contribution to the conception of the claimed invention may be named as inventors.
Europe and the UK have taken a similar position. The European Patent Office held that, under the European Patent Convention, an inventor designated in a patent application must be a human being and rejected applications naming DABUS as inventor. The UK Supreme Court has likewise confirmed that AI cannot be the inventor for UK patents.
For in-house teams, the takeaway is straightforward. The question is never “Is the AI the inventor?” The question is which people, if any, made an inventive contribution in a workflow that happens to use AI. Once the legal baseline is clear, the next question is practical: how do human inventors fit into AI-assisted discovery?
What Still Counts as Human Inventorship in AI-Assisted Drug Discovery
Patent law still hinges on a familiar concept: an inventor is the person who contributes to the conception of the claimed invention. In AI-heavy R&D, that requires looking past who clicked “run” on a model and focusing on the human thought process around what the model produces.
In practice, the people who may qualify as inventors include those who:
Define the problem, constraints, and objectives in a way that shapes the inventive concept.
Interpret and filter AI-generated outputs, identifying which results actually solve the problem.
Make non-obvious decisions about how to modify, combine, or generalize AI-proposed structures into the claimed chemical series, formulation, or method.
Translate a mass of suggestions into a specific, patentable concept captured in the claims.
By contrast, team members whose contributions are limited to routine implementation, coding infrastructure, or following instructions often do not qualify as inventors, even if their work is essential to the project.
A typical scenario illustrates the point:
An AI platform proposes thousands of small-molecule structures.
The platform ranks them based on predicted binding and ADMET properties.
A multidisciplinary team then applies domain knowledge, discards many of the top-ranked candidates, recognizes a novel scaffold with unexpected advantages, and defines a series around that scaffold that becomes the focus of the patent claims.
In that scenario, the inventors are not “the AI” or “whoever pressed start.” The inventors are the human experts who recognized, refined, and defined the inventive concept reflected in the claims. Even when inventorship is clear in theory, AI-driven R&D introduces new operational risks that can erode patent quality in practice.
Risks Specific to AI-Heavy Life Sciences Workflows
AI can create two kinds of risk around inventorship and patent quality.
First, there is a temptation to treat AI as the “discoverer” and under-document human decision-making. That shows up in lab notebooks, slide decks, and internal emails that say “the model designed this drug,” with little explanation of why the team chose certain structures, discarded others, or defined the claims in a particular way.
Second, the speed and scale of AI output can encourage very early filing on thin data sets. One analysis comparing patents from AI-native drug discovery companies to a control group found fewer in vivo studies, less biological data, and earlier filings among the AI-native group.
From an IP strategy perspective, that combination, unclear inventorship records, and thinner data, can create problems later if:
A competitor challenges inventorship or alleges that key contributors were omitted.
Due diligence for a licensing or M&A deal uncovers inconsistent stories about who did what.
Claims need to be narrowed or amended, but underlying human contributions were never properly captured.
The legal rules on correcting inventorship provide some flexibility, but they are not a substitute for disciplined front-end practice.
A Practical Playbook for In-House Counsel and R&D Leaders
Life sciences teams do not need a separate “AI patent law.” They need a practical framework for applying existing inventorship rules in AI-assisted workflows.
Here are concrete steps that help.
1. Update invention disclosure forms for AI workflows
Standard invention disclosure templates often assume traditional wet-lab R&D. For AI-assisted projects, consider adding questions like:
What AI tools or models were used, and for which steps?
Who defined the prompts, parameters, datasets, or reward functions that shaped the outputs?
Who selected, modified, or generalized the AI outputs that appear in the claims?
The goal is not to glorify prompt writing. It is to surface the people whose thinking shaped the invention.
2. Capture the human reasoning around AI outputs
Ask teams to document not just what the model produced, but why they chose specific outputs:
Which candidates were rejected, and on what scientific or strategic basis?
How did human experts refine the concept into the final claimed series, formulation, or method?
What trade-offs did they resolve that a generic optimization algorithm would not meaningfully “decide”?
Screenshots of ranked lists are less helpful than short, dated notes explaining the decision path.
3. Run structured inventorship interviews tied to the claims
Before filing, in-house or outside counsel should walk through the draft claims with the project team and ask:
Who contributed to the conception of Claim 1?
Which individuals contributed to dependent claims or alternative embodiments?
Did any contributors leave the company or move to another team?
In AI-heavy projects, it is especially important to distinguish between:
People who designed or tuned a general-purpose model that can be reused for many projects, and
People who applied that model to the particular problem and defined the claimed solution.
Only the latter are likely inventors on the specific patent application.
4. Coordinate across US, European, and UK practices
The headline rule is aligned—only humans are inventors—but procedural and documentary expectations differ by jurisdiction. By building a consistent factual narrative about who did what, and when, you reduce the risk that:
Different filings in the US, EPO, and UK name inconsistent inventor sets, or
Local counsel must reconstruct AI-related contributions from partial records years later.
A single internal “AI-assisted inventorship memo” for each major family can go a long way.
5. Train scientists and ML engineers on language and record-keeping
Team members do not need to become patent lawyers, but they do need to understand:
Why statements like “the AI invented this” are inaccurate and unhelpful.
Why do brief, contemporaneous notes about their reasoning matter for future patents and deals?
How to work with counsel early, rather than treating inventorship as a form to fill out at the end.
That training is especially important for cross-functional teams where chemists, biologists, and ML engineers are collaborating closely on shared models and pipelines.
How Hylton-Rodic Law Helps AI-Enabled Life Sciences Teams
Many of the companies using AI most aggressively in drug discovery do not have large in-house patent departments. They are expected to manage cutting-edge platforms, complex regulatory pathways, and rapidly evolving case law on AI and inventorship, often with lean internal teams.
Hylton-Rodic Law works with these innovators to:
Design invention disclosure and review processes that fit AI-assisted R&D.
Conduct disciplined inventorship analyses.
Align patent strategy with regulatory, data, and commercialization realities in drug discovery.
The technology may be new. The legal standards for inventorship are not. The companies that will win in this space are those that treat AI as a powerful tool and still keep human judgment—and human inventors—at the center of their patent strategy for AI drug discovery.
At Hylton-Rodic Law, we help life-sciences innovators navigate the legal and strategic challenges of AI-assisted R&D. To discuss AI patent inventorship or compliance with new USPTO guidance, contact our team for strategic IP counsel.
Before you file on an AI-assisted discovery or take it into diligence, make sure inventorship is defensible, and the human decision trail is documented.
Schedule a consultation with Hylton-Rodic Law to pressure-test inventorship, strengthen invention disclosures, and position your AI-enabled patent assets for licensing, partnering, or investment.
For more practical guidance on IP strategy in life sciences, chemicals, and materials innovation, subscribe to our newsletter below.
